Automated drone-based inspection of infrastructure

Masters project

Infrastructures such as bridges, electricity grids and bulk storage tanks must be inspected regularly to detect various forms of deterioration before they lead to critical faults – such as a bridge collapse, power blackouts, or environmental leaks. The frequency of inspection is often mandated by law. In many cases, the infrastructure to be inspected is located far from roadways, high above the ground, or generally difficult to access and can involve other dangers such as confined spaces, poisonous gasses, or high voltages. Rendering these environments safe for human inspectors is typically a time consuming and expensive, but necessary, procedure for the infrastructure owners.

Autonomous robotic platforms, such as drones, promise to enable safer, faster, and cheaper infrastructure inspection operations by removing the need for human inspectors to enter dangerous areas. However, there are many challenges that must first be overcome, and there is no “one size fits all” solution for every type of infrastructure and inspection operation. 

Topics of research include

• Autonomous platforms that can navigate safely and effectively even in environments that may contain obstacles or challenging localisation conditions.
• Sensor payloads with onboard AI that enable consistent, high-quality sensor data of the necessary components and/or faults even in degraded visibility conditions.
• Algorithms for digitalisation of the infrastructure, e.g. 3D reconstruction or Digital Twin, that enable fusion of multi-modal sensor data, higher-level analyses, and enhanced interpretability of the inspection data.
• Automated data analysis workflows using techniques such as Deep Learning on multi-modal sensor data to develop robust and accurate object classification, fault detection, and other high-level inspection tasks.

The Computer Vision research group at SINTEF Digital is currently working within all these problem areas (and more!) on projects including ADRIANE, SWIFTER, PILOTING, and SHEREC, amongst others. 

Research topic focus 

This thesis may cover one or more of the above topics depending on the background and interest areas of the applicant, and synergy with objectives of ongoing research.

This thesis will ideally involve research on real-world sensor data that we have access to and will produce results that are of relevance for our industrial partners in ongoing projects. Research may include field work and/or experiments with autonomous platforms.

Expected results and learning outcome

Results and outcomes may include:

• Designs for sensor payloads and/or algorithms.
• Hw/sw prototypes.
• Evaluation of prototypes on real-world data, field testing.
• Presentation/report summarising findings for stakeholders.
• Publication in academic forum.

Recommended prerequisites 

Experience/courses in image processing, computer vision, robotics, and/or processing of point cloud data will be an advantage. Familiarity with a programming language such as C++ or Python.